Identifying Key Environmental Drivers of Reach-Scale Salmonid eDNA Recovery With Random Forest

Abstract

Environmental DNA (eDNA) sampling from rivers has emerged as a promising new method for monitoring freshwater organisms of management concern. However, eDNA sampling cannot yet offer reliable estimates of a target species' abundance/biomass or confident determinations of a species' absence from a river segment. To unlock these abilities—and thereby greatly improve eDNA as a tool for management decision-making—the influence of local environmental factors on eDNA fate must be better understood. At nine river sites across the central California coast, we added a known quantity of novel eDNA (Brook Trout, Salvelinus fontinalis) and collected eDNA at sequential downstream distances for qPCR analysis. We then used random forest modeling to identify the most important environmental factors to reach-scale (≤ 200 m) sampling outcomes and characterize salmonid eDNA fate. Our final model identified six factors important to sampling outcomes, including five environmental factors (discharge, local catchment calcium oxide content, average depth of the sampling cross-section, presence of pools, and impervious cover of the watershed) and one factor regarding our experimental design (the number of qPCR technical replicates). Our results highlight the notable effects of cross-sectional area, turbulence, and catchment geology on eDNA fate, and we suggest the discharge and presence of pools as useful proxies for evaluating a site's favorability for eDNA recovery.